Abstract The TBRS*C computational model provides a mathematical implementation of the cognitive processes involved in complex span tasks. The logic of the core processes, i.e., encoding, refreshing/time-based decay, and chunking, is based on Hebbian learning, synaptic facilitation, and long-term neural plasticity, respectively. The modeling, however, takes place on a cognitive rather than a physiological level. Chunking is implemented as a process of searching for sequences of memoranda in long-term memory and recoding them as a single unit which increases the efficacy of memory maintenance. Using TBRS*C simulations, the present study investigated how chunking and central working memory processes change with expertise. Hobby musicians and music students completed a complex span task in which sequences of twelve note symbols were presented for serial recall of pitch. After the presentation of each memorandum, participants performed an unknown, notated melody on an electric piano. To manipulate the potential for chunking, we varied whether sequences of memoranda formed meaningful tonal structures (major triads) or arbitrary trichords. Hobby musicians and music students were each split up in a higher-expertise and a lower-expertise group and TBRS*C simulations were performed for each group individually. In the simulations, higher-expertise hobby musicians encoded memoranda more rapidly, invested less time in chunk search, and recognized chunks with a higher chance than lower-expertise hobby musicians. Parameter estimates for music students showed only marginal expertise differences. We conclude that expertise in the TBRS model can be conceptualized by a rapid access to long-term memory and by chunking, which leads to an increase in the opportunity and efficacy of refreshing.